Live switching 3 DVXs to tape at 24pa

D

dvxStephen

Active member
I'm doing several pieces that involve multiple cameras, both DVXs and XL2s, all running at 24pa, and destinted for a 24fps timeline in post. I've basically hit a wall when it comes to investigating firewire switchers. Our goal is to switch live, and lay the master down on DV tape, also via firewire out of the switcher -- and be able to capture the switched master into a computer for further post work.

Seems like a great idea, right? It would sure save a huge amount of capture time in post, and there'd always be the indivdual camera tapes to fall back on, say if the switch operator blew some transistions.

There are two switchers I've found that handle firewire in/out ok at 60i -- the Focus Enhancements (formerly Videonics) unit, and the DataVideo SE800. However there seems to be a wide gamut of contradictory info about whether either of these can process 24p or 24pa.

I've talked to techs at both companies, and there's interest, but no hard information. Both their web sites suggest problems with 24p re their switchers, but when I talk to the engineers, no one's sure one way or the other.

For that matter, seems to be a big lack of info on just what would be involved to make a firewire 24p switcher work right.

For example, how and where the 24p or 24pa flags are laid down?

Are they necessarily tied to individual frames throughout the footage to mark the pulldown scheme, or just head and or tail flags, etc.?

Has anyone already been down this road, that might shed some light on how to get a live firewire switch working?

Thanks!
 
This is a toughie. I've hit a wall on this before too. I think it came to a point when its too expensive for me to explore it further.
 
Although I haven't worked with NTSC cameras, or with 24p/pa for that matter, I don't think you can make it work. What I would try firs is put the camera in 24pa mode and try to record straight onto an external deck to check what the camera outputs on firewire. If you get the 2:3:3:2 with the flags (which I expect - in order to keep compatibility with the IEEE1394 protocol you have to have 60 fields), you will probably be able to use the externally recoded footage for later capture. Switching between three cameras outputting this kind of signal is impossible, because unless you have the blind luck to turn the cameras on in such a fashion that all of them follow the same 2332 pattern for each field thay record and you manage to keep themin sync without the means of an external sync gen (because you can't connect one...), as I said, switching is impossible. A possible solution would be a custom made frame buffer which would sync the three signal by means of internal memory, but I don't thing such a device exists. Somone would have to build it for you and I don't expect it to bee too cheap...
 
I think Gabriel's right, you'd probably need some sort of 24P-aware switcher that could read the incoming frames, discard the "C" frame, compensate for the 29.97/23.976 time difference, and re-encode, re-flag, (and re-generate the "C" frame) for output to a DV deck.

I don't think something like that exists. I imagine it could be built, and maybe someone will, but I don't know of anything like that...

... so unless you can exactly synchronize the cameras to start recording on the "A" frame simultaneously (which sounds impossible) I don't think you can do it and preserve the cadence.

The flags are written in the 80-bit data code that contains the timecode and user bits. Perhaps you could get a programmer to write a custom program that reads your captured master file and reconforms it to only flagged frames, but that still leaves you with the thorny issue of what if a cut happened right on the "C" frame (the split frame); that should happen roughly 20% of the time, and if you cut on the C frame there won't be any "B" or "D" frame reference to reconstruct, because half of the "B" frame would be missing.

Good luck...
 
If I'm reading you guys correctly -- and thanks so much for the detailed, thoughtful replies -- then 24p and pa, by definition, must have continous userbit-like flags that are defining the 2:3:3:2 cadence, right? (And otherwise, as I understand it, the 24p or pa has been converted by the camera -- before output to firewire or tape -- into conventional NTSC 60i that the switcher should be able to pass through ok, right?)

Hmmm. So, if the switcher passed them along, the problem would be that the three cameras aren't likely to be doing their individual 2:3:3:2 dances at the exact moment, is this correct?

If I understand this correctly, you're saying that this might screw things up during capture, as the timing definition of 2:3:3:2 would vary from camera to camera. But how is this different than capturing multiple takes, where the camera was paused, and plopping them into the same timeline? Or am I missing something?

Also, Barry, you're suggesting we might get a one frame glitch in our switcher hard cuts, if we did a hard switch and happened to hit the Cr frame, yes? That would produce a single soft frame at the cut point? I guess the odds of that are something like 1 in 5 or something like that, yes?

I can see how, during capture, a switcher dissolve in particular might create some problems -- that is, it might be somewhat confusing for the capture software that's trying to work with the 2:3:3:2 cadance, which would change in mid-stream during a dissolve.

Well, if it's likely to be a dead alley, there's one other way I could go: I could do my switched sessions 60i, and shoot the rest of the stuff 24pa, and then convert the 60i stuff to a 24 frame timeline in post. I need as much as the footage shot with maximum resolution as possible (thin line detail, etc.) -- getting past the 360 line interlaced limit -- as we've got to hold the option for film-out open, and we have some other material (that will be shot single camera, or created via animation) that's got crucial high detail content.

I'm saying this because, as I understand it, the only way to get the DVX to do thin detail and get past the 360 line limit is with the DVX's progressive settings.

So this brings me to 3 main questions:

1) Is this "best detail only in progressive mode" limitation also true with the XL2? Or can the XL2 shoot higher than 360 lines resolution at 60i?

2) The switched stuff would be either concert footage or talking heads material, where slightly less resolution might not be a deal breaker. Can you give me a feel for how much it degrades footage to convert from 60i to 24fps in post?

3) When the project's done, we can of course stay in 24fps for film-out, and DVD, but we'll also need to be converting BACK to 60i for our broadcast markets. So I'm looking at taking interview and/or concert footage from 60i to 24fps timeline for editing, and then back to 60i for output when going to broadcast. Is the 60i broadcast version likely to look noticibly less OK when viewed on a home TV than if I'd just done the whole thing in 60i?

Thanks!

Stephen
 
Hey guys,

I was intrigued by this discussion so I just ran a test. I hooked up my DVX100A to my old Sony Digital8 via firewire. I set the Sony to VTR mode and hit the record button, thus taping what was being captured by the DVX (set to 24pa - F6 factory setting). The DVX tape was not being used at all - the only record was the Sony. I ran about 10-15 seconds with some motion and some one handed still (i.e., not totally still). Then I hooked up the Sony to a 9" NTSC TV. I would say that the footage looked fine. It might have been a slight bit "beady," but good enough that if I was in your shoes, I'd do a little testing yourself. Maybe the set up you are hoping for may work.

We've always known that he conversion between 24p or pa and 60i occurs in the camera before it records to tape. But it seems the question here is whether it occurs prior to the point in the data stream where the Firewire port joins the process.

What would you expect to happen if the conversion happens between where Firewire joins the datastream and the internal 60i tape in the DVX? If it is something catastrophic, then I didn't see it. But as I said, you should check for your self.
 
I just repeated the test with outdoor footage. The clip was lousy because I was balancing the DVX in an awkward position and the Sony in the other hand. However, the actually quality was as DVX as I'm used to seeing! It was really weird having my old Sony D8 hooked up to a TV and seeing beautiful film like images coming out!!!

If this is the input the switcher receives, you're probably all set. As far as syncing the signal for the purpose of doing live transitions - well that's something I can't speak to.
 
Stephen, your understanding is pretty much accurate, yes. And regarding pausing/starting again, the camera *always* starts recording on an "A" frame, and probably always stops recording on an "E" frame (meaning, A, B, C=split, D, E, where A,B,D,E are all individual discrete pure 24p frames, and C = the split/combined B/D frame).

Any time you go interlaced, you drop the potential resolution. And if you wanted to go interlaced and then simulate 24p from that, you'll lose more... you'd end up with somewhere around 240-260 lines, vs. the potential 480 if you just shot 24p in the first place. Same with XL2.

To tell you the truth, I think the best way for you to go is to forget switching, just shoot each camera in iso, and make the edit in post. That way you can be assured of keeping and maintaining the proper 24p cadence. Using a program like the multicam plug-in scripts for Vegas can be almost as efficient as having done switching in the first place... basically you'll lose the time of capturing in the files, but once you have the files on the timeline, you play it and "switch" after the fact, generating a cutlist, and then the script automatically generates a timeline that matches what you wanted.

(and if you shoot direct-to-hard-disk, you'd even eliminate the downtime of the capture process).

I think you should forget switching, do it iso, and switch in post. You'll preserve all the resolution, you'll preserve the 24pa cadence, plus you'll have a chance to "fix" any "bad" switches you may have made during live switching...
 
Magic, thanks for doing the test!

You said, "We've always known that he conversion between 24p or pa and 60i occurs in the camera before it records to tape. But it seems the question here is whether it occurs prior to the point in the data stream where the Firewire port joins the process."

In my mind, it's pretty assured that the 24p or pa info is intact in an X100's firewire port's data stream. Otherwise it wouldn't be a featured option to record to an outside firewire camera or deck, with various choices for this in the X100's menus, including remote starting an outside deck or camera when the source DVX runs low on tape -- which the DVX manual promotes as a way of fliming continuously.

My original question was rather if it's possible to pass 24p or pa info through a switcher intact, and have this result in workable blending of multiple camera data streams at the switcher firewire output, considering that each source camera has individual 24pa pulldown subcode riding along with the video signal.

That is, would DVX or XL2 24p or pa info even make it through the switcher? If not, would there be a way to recreate the pulldown so the footage could be successfully imported into a 24fps timeline without needing to be recompressed?
 
Barry, thanks for the detailed response. Much appreciated!

You said, "...regarding pausing/starting again, the camera *always* starts recording on an "A" frame, and probably always stops recording on an "E" frame (meaning, A, B, C=split, D, E, where A,B,D,E are all individual discrete pure 24p frames, and C = the split/combined B/D frame)."

If I understand where this is leading correctly, the switcher is certain (unless someone designs internal switcher buffers to deal with all this) to initiate cuts that break the pulldown cadence, and at best, cause a bit of softening or judder for a few frames as a result, and at worst, hopelessly confuse the software in a computer during the capture process, as the cadence would change within a data stream every time the switcher's made a cut.

Am I getting the gist?

Thanks!
 
dvxStephen said:
My original question was rather if it's possible to pass 24p or pa info through a switcher intact, and have this result in workable blending of multiple camera data streams at the switcher firewire output, considering that each source camera has individual 24pa pulldown subcode riding along with the video signal.
Click to expand...

I was assuming you'd have a single DV (and therefore 60i) deck or camera that would be recording the stream that was coming through the switcher. Then, you'd connect that edited master to your NLE which would remove the pulldown on capture so you coud edit in a 24fps timeline.

I'm assuming you'd be concerned because each data stream be out of sync with the others and affect this pulldown process. Hmm, my little test doesn't answer that, does it! Well, if you try it, we'd all be interested in hearing how it goes.
 
Magic -- Your assumptin is right, but Please see Barry's replies, as he explains better than I can why this is problematic.
 
Because all of my 24pa footage stays in the realm of 60i, that's not a concern. If you think your output would be DVD or TV, this 60i route should work, for sure. But if you're going out to film - wow, that could really get tricky.
 
I'm with Barry - do your switching in post! It may be slower, but it is sure a very safe and secure approach.

I have been considering doing some of my event work with a switcher (i.e., weddings, dances, etc.), but I would not be doing 24p because the other cameras are 60i only. My motive is to spare a huge chunk of the 40-50 hours after the fact in post. But your problem seems more involved.
 
Barry, our clients have a strong interest in having a "capturable" full quality (eg:480 line) switched master, largely because there are multiple edit chains afoot, in different cities. Long story, but anyway -- if they're willing to spring for it, do you or anyone else reading this know what specifically what gear would be needed to do high quality 480 line live switching that at least equals the video quality a single DVX can pull off in 24pa mode?

I'm assuming I'm looking at renting, at minimum, a few Panasonic SDX900s. Or 800s. But do I still have the 24p issue with them, to get even these expensive cameras to produce 480 lines? Panasonic makes a switcher that has optional SDI input boards that can work with these cameras, but Michael Gee at Panansonic tells me that the SDI boards will also strip out 24p informartion. Uurg.

So am I looking at hi-def as the only viable way to switch and get 480 lines (that is, suitable to cut into a 24fps timeline for possible film-out, where 480 lines is the minimum detail level acceptable? I know zip about hi-def switchers for remote work or what they cost, if they even exist.
 
If you're shooting 24p you will still have the problem. Even if you shoot HD you'll still have the problem, as the live video signal from a VariCam is still going to be a 720/60p signal.

The easiest solution would be to shoot at 30p. If you can shoot 30p you'll get the full 480-line resolution, and you avoid all pulldown and cadence problems completely. But, you'll be stuck at 30p (no good way to get to 24p or to film or to PAL from 30P!)

I don't think you can do what you're planning. Maybe you can and there's products I'm not aware of, but I don't think it's possible.

How about this -- how about you do a live switch AND an ISO track on each camera. Let them have a switched master of a 60i-on-tape 24P image (meaning, it just passes through the switcher and gets recorded as 60i, without pulldown flags, but still retaining the look of 24p). It should look just fine, I would think. But you couldn't use it to conform for a film out. Then you also take the ISO tapes and make the version I talked about above, which would give you the option to make a frame-accurate, flag-preserved, film-transferable 24P version. It's more work, but it'd at least get you where you're trying to go.
 
I think the problem here is again the confusion between the number of scan lines in a frame and HORISONTAL resolution of a camera. An SDX900 will go beyond the 480 lines resolution. To make things clear, horisontal resolution is the maximum number of VERTICAL lines that are distinguishable on a reference screen. 400 lines means that you can point the camera at a chart that show 400 lines from left to right and you can actually count all of them in the screen. if you get a chart with more lines, what you will see is a grey picture (undistinguishable lines). This is a very vague/non-technical explanation, but it refers to your concern that your camera is not capable of reproducing 480 lines. The loss in rez in interlace mode comes from the fact that the CCDs capture only half of a frame at any given time. By the time the next half-frame is caprured, your subject moves a little, which creates a certain amount of blur. Deinterlacing will increase this blur which translates into even lower horisontal res. However, the number of lines (480) is the same.
Another issue: what you get on the firewire out is a 60i compliat signal, even if the signal carries the information for 24pa. If you tru to cut between two cameras out of sync, you will get messy cuts most of the times, probably. When you capture, there's a pull-down process hapening - you are keeping only 24 frames. When you print to tape again the pa cadence is recreated for smooth non jerky picture. As far as I know, there's no mixer capable of frame buffering 2332. The ony workaround is individual recording (on HDDs) and edit afterwards. It shouldn't tke much longer than the whole length of the event you're shooting because there's no capture time. of course, you would have to print to tape at the end.
 
Gabriel,

Thanks for weighing in. Much appreciated!

When you wrote, "An SDX900 will go beyond the 480 lines resolution. To make things clear, horisontal resolution is the maximum number of VERTICAL lines that are distinguishable on a reference screen."

Right. Recently, while evaluating Century's new Digicon filters to see if they cause a resolution loss, I shot resolution tests with both an XL2 and a DVX100a, in 24pa, fine line detail, shooting a standard EIA resolution chart. As I'm sure you're aware, this chart includes converging, finely etched parellel lines in both directions, with the resulting detail level it reveals marked at 200 to 800 lines in 100 line increments.

I blew the resulting footage up, after capture and pulldown removal, to 200% to inspect them on a 24p timeline. I found I do get over 500 lines discernable with the DVX, and a bit more with an XL2 -- IF I'm in a 24fps timeline with non-interlaced footage (pulldown removed during capture of course).

Incidentally, while the Digicon #2 doesn't visibly degrade the image sharpness when viewed on a sharp monitor (NTSC studio monitor, or even my crisp SGI non-interlaced computer LCD display) but there is SLIGHT degradation when it's blown up to 200%, say from 500 lines to 490 or so.

I haven't replicated Barry's tests, but I'll take his word for it that the DVX and XL2, when shooting 60i, limit themselves to 380 lines -- presumably to avoid creating a hornet's nest of angry customers complaining of flicker in fine lines when viewing the resulting footage on standard TVs.

So my question for you is, are you saying that the SDX900/800 cameras are capable of 480 line horizontal detail, or better, when running at 60i? In other words, is it possible to turn OFF the "thick line feature" that these cameras seem to come with - eg, so the resulting footage WOULD cause flicker if fine-lined material was later viewed on a standard CRT TV?

In my case, of course, I'd run a de-flicker filter in post to add enough blur to solve this problem for my output that might be shown on SD TVs -- but having the detail on tape would allow me to switch my SDX900 or 800 footage without incident at 60i, and have the detail I need in post for other markets. I am willing to give up the 24p timeline -- just not the detail level, since film out must remain an option.

Thanks!

Stephen Gagne
Westwood Creek Productions
Ashland, OR
 
Barry,

In your last post, you wrote, "How about this -- how about you do a live switch AND an ISO track on each camera. Let them have a switched master of a 60i-on-tape 24P image (meaning, it just passes through the switcher and gets recorded as 60i, without pulldown flags, but still retaining the look of 24p). It should look just fine, I would think. But you couldn't use it to conform for a film out...."

This is actually what I first proposed to my clients when I discovered that the "24pa through switcher" gambit was not looking so good. And you're right -- it may be the best realistic answer.

However, what's your view of doing what you suggest -- switching the 24pa footage via firewire, to produce a MiniDV switched master -- versus just switching the composite or S-Vid outputs of the DVX and XL2s?

(Video pundit David Riddle has told me that I'm far better off with using the composite signal over the s-vid, btw, if I go this route -- he says long runs of s-vid are problematic, and the quality difference is almost indistinguishable on a CRT.)

I'm assuming I'd have a SLIGHTLY lower quality signal, but beyond that I wonder -- DataVideo and Focus evidently have some struggles dealing with all the user bit info in the firewire signals of DVX and XL2; I know DataVideo is working on a fix, but not sure they've got it yet.

Do you or Gabriel know what comes out of the DVX s-vid and comp outs when one is shooting 24p or pa? Is it straight 60i? It would still have the film gamma setting, of course, so maybe that would be the cleanest, most solid thing to switch if my goal is a switched master to give clients to use for preview, EDL work, etc.

Thanks!

Stephen
 
You must log in or register to reply here.
Back
Top